To investigate systematically the natural history of visual outcome in central retinal artery occlusion (CRAO). Cohort study. At entry, 244 consecutive patients (260 eyes) with CRAO (seen consecutively from 1973 to 2000) had a detailed ocular and medical history and ocular evaluation. CRAO eyes were classified into four categories: non-arteritic (NA) CRAO (171 eyes), NA-CRAO with cilioretinal artery sparing (35), transient NA-CRAO (41), and arteritic CRAO (13). Within 7 days of onset of CRAO, initial visual acuity differed among the four CRAO types (P < .0001). In eyes with vision of counting fingers or worse, it improved in 82% of eyes with transient NA-CRAO, 67% of eyes with NA-CRAO with cilioretinal artery sparing, and 22% of eyes with NA-CRAO. Visual acuity improved primarily within the first 7 days (P < .0001). In the central 30-degree visual field, central scotoma was most common. Central visual field improved in 39% with transient NA-CRAO, 25% with NA-CRAO with cilioretinal artery sparing, and 21% with NA-CRAO. Peripheral visual field was normal in 62.9% of eyes with transient NA-CRAO and 22.1% in those with NA-CRAO. In 51.9% of eyes with NA-CRAO, the only remaining visual field was a peripheral island. Peripheral fields improved in NA-CRAO (39%) and in transient NA-CRAO (39%). Classification of CRAO is crucial for understanding differences in visual outcome. Marked improvement in visual acuity and visual field can occur without treatment and is determined by several factors. Visual field information is essential to evaluate visual disability in CRAO.